Usually costly and time-consuming, animating a 3D figure from scratch calls for sophisticated software and motion capture techniques. Cartwheel intends to make simple animations as easy as describing them, using artificial intelligence to produce a basic movement, allowing designers to concentrate on more emotive chores.
“Since I started doing it, there hasn’t been a lot of new stuff in animation,” said Jonathan Jarvis, CEO of the firm and an animator himself. With a million buttons and settings, there is a major “blank screen” issue. Sometimes you will work on something for hours before you even know what it looks like.
Animators who wish to build a scene or character can save a significant amount of time by using Cartwheel, which bypasses the initial step and moves directly to basic movements such as taking a step, swatting a fly, or sitting down. 
“We assist in generating a motion that you would achieve through a motion capture setup or keyframe animation, but at a significantly faster pace.” Simply getting it out of your thoughts and into action quickly has great value. You may then grab it and improve it, Jarvis said.
The interface is purposefully basic, designed simply as a character and text field. You can write almost anything in there, and in a minute or two, you’ll have a simple yet flexible animation that you could then export to any standard 3D editing suite.
On their website, you can even see live 3D simulations, such as this boxing one or a small guy doing a solo waltz.
According to co-founder Andrew Carr, also the company’s principal scientist, their approach is absolutely novel.
“We have labelers marking these motions and several data sources, all ethically sourced,” he said. The literature is well-known for depicting motion as a matrix, which includes postures, time, velocity, and other details. Like you would create photographs or video, you therefore link the motion matrix with a text description of the motion and undertake very conventional training on motion-language pairings.
Carr said the animations you obtain “are, on average, about 80% of the way there.” Jarvis said it may provide outstanding, professional-level performance, but sometimes it’s a fail. However, particularly in situations where many artists are working on one process and minor changes have to move up and down the line, it’s much quicker and easier than a conventional animation workflow.
The models they use might theoretically be locally hosted because they are not that large, so running costs are lower.
Carr responded, “Actually, this is rather cool.” For a video model, you are projecting 2,000,000 pixels per frame for 60 frames per second. It’s a substantial task that requires careful sorting. We can train models more quickly, and our predictions can operate on the CPU or older GPUs.
Jarvis even hinted that they could soon be able to produce new or changed animations on the fly. Though the company focuses primarily on human animations, which are by far the most necessary, things like camera motions and angles can also be clearly defined. Non-human characters are also in the pipeline.
It reminds me of how Wonder Dynamics drastically streamlined the 3D character insertion process, with an eye towards bypassing the repetitious labour of fundamental animation work rather than replacing animators or artists. Certainly, Autodesk has, at least based on its purchase of Wonder Dynamics two weeks ago, embraced AI-adjacent solutions as time savings that help its creatives concentrate more on production.
Though Cartwheel might theoretically go the same way—picketed up as a promising and strong feature that may offer an edge over rivals—it may be a bit unseemly to speculate on the eventual destiny of a business just as it’s beginning. However, it might potentially develop as a platform-agnostic tool, providing an expanding array of services to professional animators.
In any case, Accel led the $5.6 million first round of financing, which also included participation from Khosla Ventures, Human Ventures, Heretic VC, MVP Ventures, Correlation Ventures, Pelion VC, and a few angel investors.
For now, Cartwheel is content to be among the few prosumer-type applications animators depend on to do their jobs. If you registered for the beta, you may try it yourself.
“There’s this idea of artificial intelligence substituting for creative labour; as someone who creates, it’s like, no! Jarvis stated that more animation, more motion, and one person producing more results from this approach would be beneficial. Ultimately, this will elevate the bar for major production houses engaged in remarkable projects that we may not have previously considered. But there are a lot of layers between the Pixars and the people using their phones, and that’s where much of the creative work really occurs.
